There are two biostatistical components: (l) design and analysis of clinical trials and (2) evaluation of the prognostic value of novel measures of tumor proliferation, malignant phenotype, and invasiveness, all arising in the laboratory, for predicting patient response (tumor remission, survival). Preliminary analyses of novel covariates will determine distributional properties and assess possible relationships among them. These will include computation of descriptive statistics; construction of histograms, box plots and scattergrams; crosstabulation of categorical variables, and regression analyses. Evaluation of novel covariates for predicting patient response will be done primarily using logistic regression for binary endpoints and Cox proportional hazards regression for time to progression or death, including biological and genetic markers as timevarying covariates when they are measured at one or more times subsequent to baseline. These regression models will include known prognostic covariates, such as performance status, age, histology and extent of surgery, to assess the additional predictive power of novel covariates. Other survival analyses will be carried out using life tables, Kaplan-Meier curves, the generalized Wilcoxon test, the logrank test and tests for goodness-of-fit.